Leaf azimuth is not used in the estimation of light interception by a crop canopy, implying a random distribution of leaf azimuth. For a maize (Zea mays L.) crop, random leaf azimuth implies that plant orientation is random and all the leaves are in the same plane. We have tested this assumption by recording the leaf azimuths of plants grown according to different canopy patterns: plants are oriented 1) parallel to the row (± 22° 30'), 2) perpendicular to the row (± 22° 30') and 3) random. Experiments were carried out in a growth room to avoid the impact of solar path and of wind effects on leaf azimuth. Independently of the plant orientation of the young plant, less than 15 per cent of the leaves were oriented row wise, from the 9-leaf stage, which was significantly different from the initial orientation. In the crop where plants were randomly oriented, 25 per cent of them were perpendicular to the row at the 2-leaf stage and this proportion rose to 35, 45 and 70 per cent at the 7, 9 and 14-leaf stages, respectively. At the plant density used in this experiment (11 plants per m2), there was an azimuth shift from the bottom to the top of the plant, with a strong tendency to a leaf orientation perpendicular to the row for the upper leaves. When the young leaves were emerging from the whorl, the azimuth of the leaf tip changed. The final leaf azimuth was fixed for 91 per cent of the plants at 2 days before complete expansion of the leaf. At early stages, the leaf azimuth shift corresponded to a twisting of leaf sheaths, and later to a twisting of internodes. The non-random nature of leaf azimuth should be taken into consideration in future models of light interception.